danjloveless
Graduating June '26 · open to roles Exeter / London, UK BSc Computer Science · predicted 1st

Dan
Loveless✦

Final-year CS at Exeter — I build products, ship research, and run real-world ops. Co-founder of a sustainable skiwear brand distributed across 38 UK university societies, VP of the UK's largest student snowsports society (1,200+ members), author of a Rust crate with 10,000+ downloads, and lead dev of the team that won IBM's TORCS AI Race Car League.

I move comfortably between commercial, technical, and operational work — building things people use, then running the team that ships them

See selected work → danjloveless@gmail.com
01 — Selected Work
Things I've built & shipped
Eleven projects across ML research, open-source Rust, full-stack web, real-time systems, and a consumer brand. Featured first.
02 — Research
Looking inside the network
The flagship piece — my BSc dissertation, in progress.
BSc dissertation · 2025—26 · lit review 85% (top in cohort)
Mechanistic Interpretability for Hyper-Heuristic Trading Systems
A TopK Sparse Autoencoder, a Transformer meta-selector, and a five-layer causal-validation framework.

Applies a TopK SAE to a Transformer-based meta-selector for a learned trading policy, then validates the resulting features through gate analysis → sequential-split SAE R² → circuit CV R² → intervention success with BH-FDR → atom-mediation ΔKL. Five seeds; pre-registered RQs with quantitative targets. Framed against the EU AI Act and MiFID II RTS 6 Article 17.

Production-oriented artefacts: an RTS 6-mapped audit logger (19 conformance tests), a 27-field DecisionRecord schema, DuckDB audit log with retention pruning, kill-switch, Tauri/React replay dashboard, and a MultiPhaseEngine for minute-cadence deployment.

University of Exeter · Computer Science Status — Active · draft on request ↗
03 — Experience
Roles I've held
Engineering, research, and operational delivery — sometimes in the same week.
2025·
Lead AI Systems Architect
QuantFin Society, University of Exeter — leading a team building Python market-simulation and backtesting infrastructure. Defining roadmap from ingestion through strategy evaluation and risk metrics; integrating AR/CAR, CAPM, GARCH. Built an explainable-AI framework for trading-decision attribution.
Lead · current ·
2025·
VP Welfare & Opportunities
Snowsports Exeter — co-led Chill Hill 2026, a £60k, 2,000-attendee charity event that raised £10k and was featured on BBC Devon. Personally owned the sponsorship programme (30 brand partners closed from a standing start). Managing welfare + ops for an 800-person international ski trip (~£800k turnover).
Officer · current ·
2024·
Blockchain Analytics Developer
Smallcap.ai (remote) — built a real-time data pipeline (Python, SQL) processing 10,000+ daily Ethereum transactions for a client-facing risk-scoring product. Integrated Solidity smart contracts with the off-chain stack. Designed an OpenAI-based NLP layer for sentiment and risk signal extraction across on- and off-chain data.
Internship · Jun–Sep ·
2024·
Treasurer — Snowsports Exeter
Owned P&L on a £150k+ annual budget. Secured £7,360 in sponsorship including a Sport England grant. Drove 47% YoY revenue growth through tiered pricing and commercial restructuring of supplier contracts.
Officer · 1 year ·
2024·
Web Developer
Privacy Analytics Research Group, Exeter — built a Django-based platform showcasing privacy-preserving algorithms. Improved discoverability and page views by 300%.
Research · Jul–Sep ·
04 — Stack & Honors
Tools & trophies
The kit I reach for, and the moments worth keeping on the shelf.

Stack

LANGUAGES
Python · Rust · Solidity · TypeScript · C++ · SQL
ML & DATA
PyTorch · TensorFlow · scikit-learn · pandas · NumPy · snntorch · graph-tool · Hugging Face
FRONTEND
React · Next.js · Tauri · Framer Motion · vanilla JS/HTML/CSS
BACKEND & INFRA
AWS · Docker · PostgreSQL · DuckDB · Firebase · Netlify · Vercel · Git / CI-CD
SPECIALISATIONS
Mechanistic interpretability · spiking neural networks · quant finance · agent-based modelling · blockchain analytics · NN compression